ABSTRACT
This conclusion presents some closing thoughts on the key concepts discussed in the preceding chapters of this book. The book identifies a substantial number of impactable high utilizers using both count-and-cost based criteria directly from health care outcome and post-risk adjustment residuals. It demonstrates that the residual approach is highly associated with potentially preventable events. The book discusses machine learning methods to predict hospital readmission and health care expenditures, and found that our results accurately predicted high utilization. It presents both local and global model interpretation methods and temporal behaviors in health care encounters of high utilizers with unsupervised machine learning methods. Although data-driven methods could guide the development of preventive interventions, feedback from these interventions could help improve the data-driven methods. Thus, an iterative approach between data analysts and health practitioners is the best way to bridge analysis and practice and solve the high utilizer problem.
